APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES TO THE DESIGN

OF WATER RESOURCE DEVELOPMENT PROJECTS

Bi -Huei Wang
Harza Engineering Company
150 South Wacker Drive
Chicago, Illinois 60606

ABSTRACT. Designs of most major dams and spillways are based on proba-
ble maximum floods (PMF). When properly estimated, PMFs provide the
basis for confident sizing of spillway. There are, however, many fac-
tors which affect the estimation of a PMF. In analyzing each of these
factors, certain assumptions have to be made and subjective judgements
exercised. Depending on these assumptions and judgements, reSUlting PMF
estimates may be quite different. This paper provides a brief summary
of current practices in PMF estimation and application, and offers some
guidelines for such practices. It also includes a comparison and dis-
cussion of PMF and risk-based analysis for spillway design.

1. INTRODUCTION

The main application of probable maximum flood (PMF) estimates to the

design of water resource development projects is when the projects in-
volve the construction of major dams and spillways. Any dam constructed
on a natural stream requires a spillway to safely pass flood flows down-
stream. Unfortunately, it can never be certain how large a flood a dam
will experience during its useful life. Therefore, any spillway will
likely be designed either too large or too small.
If the spillway is designed to pass floods larger than those which
could possibly occur, a large part of the investment may be wasted in
providing excess capacity which will never be used. On the other hand,
if it is designed to pass only .floods smaller than those which could
possibly occur, there will be a risk of failure of the dam and associ-
ated consequences.
Ideally, the selection of spillway capacity should be based on
comprehensive economic, social, and political considerations, as well as
technical criteria of hydrology, hydraulics and structural design. In
practice, however, a criterion commonly used is to design the spillway
for the PMF if it involves a high dam impounding a large amount of
water, or where failure of the dam is likely to cause loss of human
lives.

23
v. P. Singh fed.), Application of Frequency and Risk in Water Resources, 23-33.
1987 by D. Reidel Publishing Company.
24 B. H. WANG

The PMF as defined by the U.S. Army Corps of Engineers (1979) is

"the flood that may be expected from the most severe combination of
critical meteorological and hydrologic conditions that are reasonably
possible in the region". It is a hypothetical flood estimated by trans-
forming a probable maximum precipitation (PMP).

In the following section, a brief summary of current practices in

the estimation of PMP and its time and areal distributions is given. It
is followed by a section on how an estimated PMP is transformed to a PMF
which, in turn, is followed by a section on application of the PMF. The
final section is devoted to a comparison and discussion of PMF and risk-
based analysis for spillway design.

2. ESTIMATION OF PMP

PMP was once known as maximum possible precipitation (MPP). The main
reason for the change from MPP to PMP was that MPP carried a stronger
implication of a physical upper limit of precipitation than does PMP.
PMP is preferred because of uncertainties surrounding any estimate of
maximum precipitation.
In a publication by the World Meteorological Organization (1973),
PMP is defined conceptually as "the theoretically greatest depth of
precipitation for a given duration that is phYSically possible over a
particular drainage basin at a particular time of year". Operationally,
however, it is the estimated precipitation for a given duration, drain-
age area, and time of year for which th~re is virtually no risk of being
exceeded, assuming that there is no long-term climatic trend.

2.1. Critical Duration

The magnitude and intensity of a PMP varies with duration. The duration
of the PMP that causes the most critical flood at a damsite is the cri-
tical duration. In general, the critical duration is short for a small
basin and long for a large basin. The flood regulating capacity of
reservoirs located at or upstream of a project site also affects the
length of critical duration.
To determine the critical duration, inflow hydrographs resulting
from PMP of various durations should be derived and routed through the
proposed reservoir. The duration of the PMP that causes the maximum
flood surcharge and outflow is the critical duration. In practice,
however, to avoid the need for estimating PMPs of various durations, the
critical duration is often estimated based on the size of the drainage
area and the flood storage capacity of all reservoirs located at and
upstream from the proposed site.
For a small drainage area and a small reservoir storage capacity, a
short duration of six hours or less may be critical. On the other hand,
for a large drainage area and a large reservoir capacity, a long dura-
tion of three days or more may be critical.
The general guideline for determining the critical duration is that
it should be at least equivalent to the time of concentration of the
APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES 25

basin, and should be longer when a reservoir of substantial storage

capacity is considered. In case of any doubt, a number of durations
should be examined to determine the most critical one.

2.2. Type of Storm

For the purpose of estimating the PMP for a specific basin, rain storms
can be classified as "general-storms" or "local-storms". A general-
storm normally lasts more than 6 hours. It is associated with a major
synoptic weather system and generally produces precipitation over an
area in excess of 500 mi 2 (1,300 km2 ). A "local-storm" seldom lasts
more than 6 hours. It is commonly referred to as a thunder storm and
produces precipitation over an area of less than 500 mi 2 (1,300 km 2 ).
The magnitude and time distribution of general- and local-storm
PMPs are generally quite different. For a project basin with a critical
duration of 6 hours or less, local-storm PMP generally would produce the
PMF. For a critical duration of more than 24 hours, a general-storm
normally would be more critical. For a critical duration of 6 to 24
hours, both types of storms would need to be analyzed to determine
whether the local- or general-storm PMP would be more critical.

2.3. Generalized Estimates

For most of the United States, generalized PMPs have been estimated and
published by the National Weather Service (previously the Weather
Bureau) in collaboration with other agencies such as the Army Corps of
Engineers, Bureau of Reclamation, the Tennessee Valley Authority, and
the Soil Conservation Service. These publications are contained in a
series of Hydrometeorological Reports, or Technical Papers, available
from the Government Printing Office or the National Weather Service.
Using generalized PMP available in these publications, PMPs of
various durations can be estimated for a storm area of any size and at
any location. Based on these estimates, a synthetic PMP isohyetal map
can be constructed and superimposed on the basin maps of the same scale.
A step-by-step procedure for applying the generalized PMP estimates
generally is included in the publication containing the estimates.

2.4. Site Specific Studies

There are basins of unique topographic characteristics where generalized

PMP estimates cannot be properly applied. Under such conditions, or in
areas where generalized PMP estimates are not available, a site specific
study is needed (Riedel, Wang and Diebel, 1982). Basically, this
involves the transposition of major historical storms from where they
occurred to the project basin and the adjustment of the recorded rain-
fall for maximum moisture, topographic effects, and possibly other fac-
tors such as wind speed.
An important factor that affects the result of any site specific
PMP study is the number of major storms considered in the study. If the
region from which the recorded storms are transposed is too small or if
the period of storm rainfall records is too short, the site specific
 B.H.WANO

study may result in an under-estimated PMP. Therefore, identifying the

region which is considered to be meteorologically homogeneous with the
project basin is extremely important. This may require an extensive
study of meteorology of major rainstorms in a large region.
Another factor that greatly affects a site specific study is
adjustment of the transposed storm for PMP. This is especially true for
orographic regions where topographic effects are important. For such
regions, an adjustment factor is determined to account for terrain
differences. This is often accomplished by using an index map, such as
the 100-year 24-hour precipitation map or seasonal isohyeta1 map, to
compute the ratio of the index rainfalls in the project basin and in the
area where the storm occurred. The resulting ratio is multiplied by the
moisture maximization factor to obtain the combined adjustment factor
(Wang, 1984).
The time distribution of PMP for a site specific study is also
important. This can be determined in several ways depending on the
degree of conservatism required for the PMF. The most conservative
approach is to construct depth-duration curves of major storms in the
region, draw an envelope curve, read incremental rainfall from the curve
and arrange them in such a way so as to yield the most critical flood.
A less critical but still conservative way is to construct rainfall mass
curves of major storms in the region, and draw a near envelope curve
which nearly envelopes the low values in the low end of the curves and
high values in the high end of the curves (Wang, 1984). A less conser-
vative approach would be to use the actual distribution of the storm
which, after transposition and maximization, yielded the PMP estimate.

2.5. Snowmelt

For a project basin where snowmelt contributes significantly to major

floods, critical combinations of rainfall and snowmelt must be consid-
ered. In general, the two combinations listed below are investigated:
(1) PMP during the snowmelt season plus snowmelt of a 50- to 100-
year recurrence interval.
(2) Probable maximum snowmelt plus a 50- to 100-year rainfall
during the snowmelt season.
These and possibly other combinations of rainfall and snowmelt are
then compared with all-season PMP to select the most critical combina-
tion of events for estimating the PMF (Wang, 1986).

2.6. Antecedent and Subsequent Storms

A cyclonic storm could precede or follow another cyclonic storm within a

3- to 5-day period, depending on the season of the year and the meteoro-
logical characteristics of the region. Sequential thunderstorms could
occur with even shorter periods between the storms. The antecedent and
subsequent storms mayor may not be important in the PMF analysis
depending on their relative magnitudes with respect to the discharge
capacity of the proposed reservoir at normal maximum pool.
In general, if the discharge capacity at the normal maximum pool is
larger than the peak flow caused by the antecedent storm, it would not
APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES 27

be necessary to consider the antecedent storm. Consideration of a sub-

sequent storm also would not be necessary if the discharge capacity is
large enough to return the high reservoir level caused by the PMP back
to the normal maximum pool before the subsequent storm.
On the other hand, if the discharge capacity of the proposed reser-
voir is so small that the antecedent flood cannot be discharged before
the PMP occurs, or if the subsequent storm would add to the maximum
flood surcharge caused by the PMP, the antecedent and/or subsequent
storms will have to be considered in determining the PMF.

3. TRANSFORMATION OF PMP TO PMF

Transformation of a PMP to a PMF is generally accomplished by the unit

hydrograph method or by a simulation model (Wang and Jawed, 1986). The
first method is more conventional, relatively simple, and generally
requires less data and effort. The second method is more sophisticated
and comprehensive but requires more data to justify its employment. In
applying either method, an important step is to have a careful recon-
naissance of the project basin and to divide the basin into a number of
sub-basins so that areal variation of rainfall and basin characteristics
can be properly accounted for. The number of sub-basins into which the
project basin is divided depends on the size of the basin, the PMP pat-
tern, and the basin characteristics such as topography, vegetative
cover, and soil. The general guideline is to divide the basin in such a
way that areal variation in rainfall, soil, vegetative cover, and topo-
graphy in each sub-basin would be as small as practicable. For most
cases, five to fifteen sub-basins are required.

3.1. Unit Hydrograph Method

Transformation of a PMP to a PMF by the unit hydrograph method requires

estimation of retention losses, derivation of unit hydrographs, convolu-
tion of rainfall excesses with unit hydrographs, estimation of base
flow, and routing of floods through the stream and reservoir system as
discussed below.

3.1.1. Retention losses. Not all of the PMP runs off during the flood
period. Some is intercepted by foliage before it reaches the ground,
some is retained in surface depressions, some is lost due to evapotrans-
piration, and some infiltrates into the ground. These losses are gener-
ally grouped into a single term called retention losses.
Estimation of retention losses generally is accomplished by one of
two methods; the infiltration approach and the runoff curve method.
The infiltration approach follows the concept of representing the re-
tention losses with a fixed amount of initial retention followed by a
continuous retention at a constant rate. The initial retention may
range from a few tenths of an inch (a few millimeters) during wet sea-
sons to over 4 inches (100 millimeters) during dry seasons, depending
primarily on the soil and vegetative cover of the basin. However, major
storms generally occur in wet season during which the soil is wet or
28 B. H. WANG

saturated. Therefore, initial retention is generally assumed to be

negligible.
The continuous retention rate represents the infiltration capacity
when the soil becomes saturated. It is controlled by the capability of
lower soil horizons to transmit water downward, and may be determined by
a trial and error process if rainfall and corresponding runoff data are
available. In the absence of sufficient rainfall-runoff records, empir-
ical retention rates are often used based on hydrologic soil groups
(U.S. Bureau of Reclamation, 1973).
The runoff curve method uses soil and ground cover data for estima-
tion of retention losses. Various tables and charts have been developed
by the u.S. Soil Conservation Service to determine runoff curve numbers
based on which direct runoff due to rainfall is estimated (U.S. Bureau
of Reclamation, 1973).

3.1.2. Unit hydrographs. A unit hydrograph is the hydrograph resulting

from one unit of rainfall excess over a specified duration. Generally
accepted methods for deriving unit hydrographs include the direct method
(Linsley, Jr., Kohler and Paulhus, 1975), the dimensionless graph-lag
curve method (U.S. Bureau of Reclamation, 1973) and synthetic methods.
The direct method is applicable when there is a stream gaging sta-
tion at or near the damsite and the drainage area upstream from the site
is so small that areal variations of the storm rainfall, and basin char-
acteristics such as topography, soils, and vegetative cover, can be
ignored.
The dimensionless graph-lag curve method may be used to derive a
unit hydrograph for any point of interest in a basin where flood hydro-
graphs have been recorded at various gaging stations in the basin or its
vicinity. The points of interest, including the damaite, mayor may not
coincide with the gaging sites.
For ungaged basins where no recorded flood hydrographs are avail-
able for analysis, synthetic methods such as the Snyder's method
(Linsley, Jr., Koaler and Paulhus, 1975), the U.S. Soil Conservation
Service's method (U.S. Soil Conservation Service, 1972), and Clark's
method (Clark, 1945) can be used. The procedure for applying each of
these method is described in the respective references.

3.1.3. Convolution. The direct runoff hydrograph resulting from a storm

is derived by convolution of the incremental rainfall excesses with the
appropriate unit hydrograph. Briefly, for each sub-basin, the direct
runoff resulting from the first incremental rainfall excess is calcul-
ated by multiplying it with the unit hydrograph ordinates. The direct
runoff resulting from the second incremental rainfall excess is then
computed similarly and added to that resulting from the first incremen-
tal rainfall excess with the time lagged by one time interval. This
procedure is continued until all incremental rainfall excesses are pro-
cessed. The time increment for the rainfall excess is selected to be
equal to the unit duration of the unit hydrograph.

3.1.4. Base flow. Base flow is that portion of streamflow derived from
groundwiterstOrage. Its contribution to the PMF peak discharge and
APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES 29

volume is generally small, and a simplified estimate of base flow nor-

mally would not cause appreciable difference in the PMF estimate. A
time-saving approach which is often adopted is to use the minimum daily
flow during the month of highest flow in the period of record to approx-
imate the base flow.
3.1.5. ~ and reservoir routing. When a basin upstream of a damsite
is treated as one unit, the computed hydrograph is the PMF inflow hydro-
graph. When a basin is divided into a number of sub-basins, the hydro-
graph derived for the most upstream sub-basin is routed through the
stream and reservoirs, if any, to the outlet of the second most upstream
sub-basin. The routed hydrograph is then combined with that from the
second most upstream sub-basin. This procedure is continued until all
sub-basins are processed. The resulting hydrograph is the PMF inflow
hydrograph.

3.2. Simulation Techniques

A simulation model generally consists of a number of mathematical equa-

tions each describing a component of the hydrologic cycle such as pre-
cipitation, interception, evaporation, transpiration, infiltration,
percolation and runoff. These equations are linked by the mass continu-
ity equation and sometimes also by the momentum equation, and are pro-
grammed on an electronic computer for rapid solution.
There are a large number of simulation models that have been devel-
oped by various organizations. Some of the models most commonly used
include the HEC-1 and HEC-5 models developed by the U.S. Army Corps of
Engineers, The Hydrologic Engineering Center (1981, 1979), the River
Forecast System developed by the U.S. National Weather Service (1972),
and the Stanford Watershed Model developed at Stanford University
(Crawford and Linsley Jr. 1966). Any of these and other models can be
used to transform a PMP to a PMF provided that sufficient data are
available to properly calibrate the model.
On the other hand, for basins where sufficient data are not avail-
able, overly anxious employment of a simulation model can lead to erro-
neous results. Therefore, selection of the more conventional unit
hydrograph method or the more recently developed simulation techniques
should be based on a careful evaluation of available data.

4, APPLICATION OF PMF ESTIMATES

Before a PMF estimate is adopted for design purposes, its reasonableness

is often investigated further by comparing the PMP and PMF estimates
with recorded rainfalls and floods.
Observed rainfall data, such as those given in the "Manual of
Probable Maximum Precipitation" by' the World Meteorological Organization
(1973, pp. 175-179), can be used to compare PMP estimates for their
reasonableness. However, geographical, meteorological, and topographic
similar! ties between the study area and the area used for comparison
should be carefully evaluated and taken into consideration. For
example, an extreme rainfall, such 8S the September 3-7, 1950 storm that
30 B.H. WANG

occurred at Yankeetown, Florida, potentially could occur at other loca-

tions in southeastern United States but not in the Rocky Mountain region
of Colorado.
Flood records are generally shorter than precipitation records, and
extreme floods mayor may not have occurred during the period of record.
Therefore, a comparison of PMF with historical floods is more difficult
and needs to be made with great caution. The procedure outlined below
has been useful in making comparisons for some projects.
1. Conduct a thorough search of historical records on major
floods in the region. These records may be of flood magni-
tudes, flood marks, and/or flood damages published in govern-
ment or private documents and newspapers.
2. Perform a field investigation and interview of long-time local
residents for their recollection of the floods. Questions to
ask the residents include time, date, and year of occurrence
of major floods, flood marks and their relative magnitudes
with respect to the largest flood of known magnitude.
3. Estimate the magnitudes of these floods and determine the
maximum flood of record for each location.
4. Plot the floods of record against their respective drainage
areas and draw an envelope curve.
S. Compare the flood of record from the envelope curve with the
PMF estimate. In general, the ratio between the PMF and the
flood of record may be as small as two to three in tropical
areas frequently invaded by typhoons or hurricanes but it may
be very large in arid or semi-arid areas where heavy storms
are scarce but do occur.

PMF estimates reSUlting from the analyses and comparisons described

above often lead to large spillways which might be designed to discharge
flows much larger than that which would occur under natural conditions.
This, however, can be avoided by designing the dam and spillway to allow
some flood surcharge in the reservoir. Generally, the PMF inflow hydro-
graph is routed through the proposed reservoir for a number of spillway
configurations to determine the maximum flood surcharges in the reser-
voir and the maximum discharges through the spillway and other outlet
facilities. Results of such routings are then used to select the opti-
mum height of the dam and best design of the spillway based on economic
and other considerations.
Under situations in which construction of a regular spillway to
pass the entire PMF requires excessive investments, a smaller regular
spillway supplemented by one or more fuse-plug spillways is occasionally
used to reduce the total cost. Each fuse-plug spillway is designed to
be activated when the reservoir level reaches a predetermined elevation.
The regular and fuse-plug spillways, however, are designed such that,
under no circumstance, will the total discharge from the reservoir
exceeds the peak inflow. This often requires the construction of two or
more fuse-plugs designed to be activated at different elevations.
In case of a pumped-storage project, water is released from the
upper reservoir through the power house to the lower reservoir to gener-
ate electricity during the periods of peak demands. This water is
APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES 31

pumped back to the upper reservoir during the off-peak hours. For such
a project, special attention is required in applying the PMF estimate.
This is because the PMF can arrive at any time in the pumping-generating
cycle, and depending on when in the cycle the PMF arrives, the lower, or
the upper, reservoir surface can rise rapidly and over-top the dam. In
such cases, a flood operation plan has to be formulated to interrupt the
normal operation cycle when the incoming flood exceeds a certain magni-
tude.

s. DISCUSSION AND CONCLUSION

There are two basic approaches for determining design flooda for maj or
dams and spillways. One approach is to use the PMF and the other
approach is to use a flood of selected recurrence interval based on a
risk analysis.
The first approach is based on the philosophy that dam failures
should be prevented at any cost if such failures would cause loss of
human lives. This approach does not directly consider optimum alloca-
tion of national resources but it offers the basis on which the engi-
neering profession has the most confidence for sizing spillways. It
appears to meet a stsndsrd of reasonable care as demonstrated by the
performance of dams over the past several decades (U.S. National
Research Council, 1985). It also minimizes dam owner's liability on
damages due to failure of their dams. This is becoming increasingly
important in view of recent court cases which indicate a trend of victim
compensation for damage caused by dam failures.
The second approach is based on the theory that society cannot
afford the cost of preventing all dams from failure and that the assign-
ment of an infinite value to human life is wasteful and contrary to
common practices on expenditures for safety in other phases of human
society. This approach allows the incorporation of risk analysis in
selecting the design floods and, thus, avoid overdesign of dams and
spillways. The difficulty of a credible risk analysis, however, is
tremendous. This is because most major dams and spillways would have to
be designed for a flood having a recurrence interval of several thousand
years or more. Any estimstion of such a flood is subject to great
uncertainty. The estimation of cost of risks also requires the determi-
nation of a flood damage function. Accurate determination of such a
relationship between damage and flood magnitude is often difficult,
especially considering the general reluctance in assigning monetary
value to human life. Furthermore, a risk analysis considered valid
today may be nullified in a short period of 10 or 20 years by changes in
downstream development and other factors included in the risk analysis.
In a recent publication by the U.S. National Research Council
(1985), an inventory of current practices on dam safety against extreme
floods is given. This inventory shows considerable diversity in
approach by various federal, state, and local government agencies and
privately owned companies. However, there is a fair concenSUB on the
spillway requirements for large, high hazard dams. This concensus is to
design a new large, high hazard dam for the PMF. No similar concensus
32 B.H. WANG

exists for existing dams of high hazard potential or new and existing
dams of intermediate and low hazard potentials. For such dams, floods
ranging from a PMF to a much smaller flood based on a risk or probabil-
ity analysis are used for design purposes depending on the individual
situation of the dam.
In any case, the utilization of PMFs for design of new major dams
will likely continue. It is also likely that employment of a risk-based
analysis in the process of selecting or evaluating the design flood will
increase. Results of the risk-based analysis, however, will likely be
used as supplementary information in the design of new large high hazard
dams although they may be used as the sole or primary basis in the
design of new, smaller, intermediate and low hazard dams. Risk-based
analysis is also likely to find increased application in the selection
of alternative measures for upgrading spillway capacities of existing
dams of intermediate- and high- hazard potentials.

REFERENCES

(a) Books.

Linsley, Jr., R.K., Kohler M.A. and Paulhus, J.L.H. 1975.

Hydrology for ~ngineers, 2nd ed., McGraw-Hill Book Co., New
York, N.Y.

U.S. Bureau of Reclamation 1973. Design of Small Dams, 2nd ed.,

Washington, D.C.

(b) Journals and Periodicals.

Clark, C.O. 1945. 'Storage and Unit Hydrograph', ASCE Transac-

tions, Vol. 110, pp. 1419-1488.

Wang, B.H. 1984. 'Estimation of Probable Maximum Precipitation:

Case Studies', ASCE Journal of Hydraulic Engineering,
Vol. 1l0, No. 1l0, pp. 1457-1472.

Wang, B.H. and Jawed, K., 1986. 'Transformation of Probable Maxi-

mum Precipitation to Probable Maximum Flood, Case Studies',
ASeE Journal of Hydraulic Engineering, Vol. 112, No.7,
pp. 547-561.

(c) Edited Symposia.

Riedel, J.T., Wang, B.H. and Diebel, J.L. 1982. 'Site Specific
Probable Maximum Precipitation Estimates, Upper South Platte
River Basin Colorado', in: A.I. Johnson and R.A. Clark
(Editors), Proceedings of the International Symposium on
Hydrometeorology, American Water Resources Association,
Bethesda, Md. pp. 517-522.
APPLICATION OF PROBABLE MAXIMUM FLOOD ESTIMATES 11

(d) Technical Reports.

Crawford, N.H. and Linsley Jr., R.K. 1966. 'Digital simulation in

Hydrology: Stanford Watershed Model IV', Stanford University,
Department of Civil Engineering, Technical Report 39, Palo
Alto, Calif.

U.S. Army Corps of Engineers 1979. 'National Program for Inspec-

tion of Non-Federal Dams', ER 1110-106, Appendix D, Office of
the Chief of Engineers, Washington, D.C.

U.S. Army Corps of Engineers, The Hydrologic Engineering Center

1979. 'HEC-5, Simulation of Flood Control and Conservation
Systems, Users Manual, Generalized Computer Program 723-500',
Davis, Calif.
U.S. Army )orp of Engineers, The Hydrologic Engineering Center
1981. 'REC-l, Flood Hydrograph Package, Users Manual, Gener-
alized Computer Program 723-X6-L2010', Davis, Calif.

U.S. National Research Council 1985. 'Safety of Dams, Flood and

Earthquake Criteria', National Academy Press, Washington,
D.C.

U.S. National Weather Service 1972. 'National Weather Service

River Forecast System Forecast Procedures', NOAA Technical
Memorandum NWS Hydro-14, Silver Spring, Md.

Wang, B.H. 1986. 'Probable Maximum Flood and Its Application',

Harza Engineering Company, Chicago, II.

World Meteorological Organization 1973. 'Manual for Estimation of

Probable Maximum Precipitation', WMO No. 332, Geneva,
Switzerland.